Introduction
Q:How brain regions,
implicated in visual
WM, modulate their
activity based on the
attention demands?
Question
Use the HCP dataset, and utilize generalized linear
model (GLM) to compare the BOLD signal under two
conditions (via SPM package), one with high working
memory demand (n-back) and another with low
working memory demand (0-back).
We compare both conditions in terms of changes of
activity and in functional connectivity.
What we have done
Working memory task in HCP dataset
√ 0-back, 2-back
A
X
A
C
B
2-back
A
X
A
C
B
X
0-back
X Tools, Body, Faces, Scenes
Subject1
Subject2
Subject3
...
Region 1
Region 2
...
t
What the data looks like?
@
C1
C2
C3
…
C11
Model
params
White noise
Level 1
Brain “activations” (fitted coefficients)
for subjects, regions
C1s: S1, S2, S3, ... Sn
C2s: S1, S2, S3, ... Sn
Level 2
T-scores for each region
(test for related samples,
with Bonferroni correction)
Design matrix
BOLD for single person, region
Linear model
Regions which are activated significantly
more active in 2-back more active in 0-back
Auditory
Visual 2
Left Right
Language
Default
Auditory,
Language
Orbito-
Affectional
Dorsal attention
T-scores (significant)
T-scores (significant)
more active in 2-back more active in 0-back
Visual 2
Top Bottom
Auditory
Language
Default
Auditory,
Language
Dorsal
attention
Functional connectivity
Rearrange
Task-specific
correlation
between
regions
Average by subject,
Filter (with
threshold)
2-Back 0-Back
Parietal Lobe
Frontal Lobe
Top View
Left View
Gratitudes
We want to thank our project mentor, Adeel Razi (Monash University, Australia) for his
guidance!
We would also like to thank our TA and project TA — Madhukar Dwivedi
and Raul Hernandez for their daily help!
Thank you for your attention!
Watery camels, by Zekai Xu
